Here's a number that should make hospital CFOs wince: In Medicare Advantage alone, health insurers processed nearly 53 million prior authorization decisions in 2024. Tens of thousands get denied daily. And yet, barely 11.5% ever get appealed.
The economics behind that hesitation are brutal. Of the small fraction of denials that providers actually contest, more than 80% get overturned—fully or partially. It's a staggering reversal rate, one that suggests the initial denials were questionable to begin with. But fighting them takes time, expertise, and administrative stamina that most practices simply don't have. The gap between what could be recovered and what actually is has persisted for years, a market failure sustained almost entirely by friction.
Enter Insurf, a two-person outfit fresh out of Y Combinator, which is now trying to automate the whole messy business. This month the company launched Inveto, an AI platform designed to convert denied healthcare claims into physician-attested appeals—complete with citations, evidence snapshots, and deadline tracking.
Whether two founders can crack a problem this entrenched, in a market this crowded, is the open question.
Attestation as a Feature
Co-founders Chang Lu and Bryan Chung aren't selling speed, exactly. They're selling accountability—or at least, that's the pitch. Inveto operates through what the company describes as a six-step pipeline: intake of the denial, drafting an appeal with source citations, physician review and formal attestation, staff-controlled submission, follow-up, and outcome logging.
The physician attestation step is treated as a distinct gate. No appeal gets labeled "physician-attested" until a clinician actually signs off. In theory, that addresses one of the thorniest compliance issues in the appeals space: who takes responsibility when an AI generates the argument?
The platform pulls from a range of inputs—denial letters, payer policies, medical necessity guidelines, LCD and NCD references, clinical notes. Every claim in the draft gets tagged with a source identifier. Evidence is archived. The company has even published its reporting methodology online, specifying how it will count cases, suppress small sample sizes, and separate payer decisions from actual revenue collected. That level of methodological transparency—before a single customer outcome has been made public—is rare in the revenue cycle automation world.
Most vendors in this space lead with conversion rates and dollar recovery. Insurf, by contrast, is running its public demo on synthetic data. Its status page notes plainly that production-level protected health information remains disabled until business associate agreements and security protocols are locked down. SOC 2 compliance is listed as in progress, no certification issued yet.
It's a cautious posture, perhaps more cautious than the market typically rewards.
Regulatory Tailwinds, and Lawsuits
The timing, at least, seems deliberate. Regulatory and legal pressure around prior authorization denials has been building for years, and it's starting to boil over.
CMS finalized an interoperability and prior authorization rule requiring certain payers to publicly report authorization metrics starting in 2026—a transparency measure aimed squarely at the kind of opaque decision-making that produces an 80% overturn rate. API requirements for provider access to prior authorization data are set to take effect soon, forcing payers to open up systems they've historically kept sealed.
Then there's the litigation. ProPublica's 2023 investigation into Cigna's PXDX system—a process that allegedly allowed doctors to deny claims in seconds without reviewing patient files—ignited a firestorm. Ongoing litigation and court orders have forced increased scrutiny of algorithmic denial practices, with courts no longer inclined to treat AI-driven denial engines as black boxes.
The Kaiser Family Foundation analysis that revealed the 80.7% overturn figure didn't just expose an appeals gap. It quantified, case by case, the money left on the table by hospitals and specialty practices too stretched to fight back. For a startup like Insurf, that's the addressable market.
A Crowded Field
Except Insurf isn't alone in spotting the opportunity. The competitive landscape has gotten dense fast.
PayerWatch offers AI-assisted appeals drafting for hospitals. Recoupr targets therapy practices. Appealio and ClaimVise promise end-to-end automation. Intelliga Health, another Y Combinator company, is working the prior authorization angle. Then there's MedAppeals, AppealGenius, Mulligan AI—the list of startups positioning around denial management has lengthened considerably over the past year or so.
What Insurf says differentiates it is rigor. The methodology page on the company's website is unusually specific about what it will and won't claim: no assertions of payer misconduct without source backing, no extrapolation from pilot data to market-wide trends, no publication of patient-level outcomes. The company promises what it calls an "accountability contract" for pilot customers—worked-or-explained coverage of eligible cases, fee credits if deadlines are missed, clear separation between payer decisions and actual cash collected.
It's the kind of thing that might reassure a hospital compliance officer. Whether it's enough to win deals when competitors are moving faster and promising bigger numbers is less clear. Insurf has no publicly named pilot customers. No pricing structure. No throughput benchmarks. Its status page describes Inveto as being in "private pilot readiness," whatever that means in practical terms.
The Longer Game
Inveto, though, is just the opening move. Every resolved appeal feeds into what Insurf is calling the "coverage-decision graph"—a dataset that maps which treatments get denied, by which payers, under which clinical circumstances. That graph, in turn, is meant to power the company's second product: Surely, a cost simulation engine for employers and brokers that models 12-month health plan expenses by layering denial risk on top of premiums, deductibles, and out-of-pocket maximums.
It's an ambitious arc. Turn denials into structured data, then sell that intelligence back to the market as predictive fuel for benefits planning. Whether it works depends on whether Insurf can build the graph fast enough to matter—and whether anyone will trust a dataset assembled by a two-person startup incorporated just months ago.
Y Combinator's standard seed investment is typically $125,000, though the company hasn't publicly confirmed specific funding terms. The team is working out of San Francisco. Delaware incorporation. Standard-issue startup infrastructure.
The real test is execution. Can two founders move quickly enough in a space where dozens of competitors are chasing the same denied claims? Does methodological discipline matter more than time-to-market when a hospital revenue cycle director is choosing which vendor gets access to sensitive claims data?
The 80.7% overturn rate isn't going anywhere—it's a structural feature of how American healthcare rations care through administrative burden. The race now is to see who captures it, and whether the winners will be the ones who moved fast or the ones who moved carefully.
For Insurf, the bet is that rigor wins. The market will render its verdict soon enough.
